convez.ai
MCP ServerIntegração & APIsIntermediário
pipeworx-io

Pipeworx MCP para Cursor

pipeworx-io/cursor-plugin

Acesso a 3.300+ ferramentas de dados ao vivo via roteamento inteligente em Cursor

Ver no GitHub
InicialPluginScore · de 0 a 100

Server MCP que expõe 3.300+ ferramentas de dados ao vivo de 750+ fontes (SEC, FDA, FRED, Census, EPA, USPTO, ATTOM, clima e mais) através de um gateway inteligente. Em vez de carregar 3.300+ esquemas de ferramentas, usa roteamento natural em tempo de execução com apenas ~26 meta-ferramentas. Perfeito para pesquisa financeira, análise de propriedades, dados governamentais e biotech.

Instalação

$npx convez-ai pipeworx-io/cursor-plugin

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
50%
Último commit
há 1 mês
Versão
v0.2.3
Issues abertas
0
Contribuidores
2

Casos de uso

  • Pesquisa financeira (SEC filings, histórico de cotações, análises fundamentalistas)
  • Inteligência de patentes e propriedade intelectual (USPTO, search de propriedades ATTOM)
  • Análise de dados governamentais (Census, FRED, USAspending, EPA)
  • Pesquisa biofarmacêutica (calendários de catalisadores, pipelines clínicos)
  • Análise de mercado imobiliário (valores medianos, demanda por metros, dados de aluguel)

Veredito

Solução robusta e bem-pensada para acesso a ecossistema massivo de dados públicos sem inflacionar o context window. Arquitetura de meta-ferramentas é elegante; tier gratuito (50 chamadas/dia) e documentação clara facilitam adoção. Recomendado para pesquisadores, analistas quantitativos e aplicações que demandem inteligência de múltiplas fontes oficiais.

README

A tradução para o português deste README ainda está na fila.

Pipeworx for Cursor

Give Cursor one MCP that reaches 3,300+ live-data tools across 750+ sources — SEC filings, USPTO patents, FRED, Census, FDA, EPA, USAspending, Polymarket, Zillow, weather, and 740+ more — without loading 3,300+ tool schemas into your context window every turn.

Install

From the Cursor Marketplace:

cursor.com/marketplace → search "pipeworx" → Install

Or one-click via deeplink:

cursor://anysphere.cursor-deeplink/plugin/install?name=pipeworx

Try it

After install, ask Cursor things like:

AskWhat it triggers
"What just happened to Apple?"sec_8k_recent → SEC 8-K events classified by severity
"Spread between Polymarket and Kalshi on the next Fed decision?"polymarket_kalshi_spread → live cross-venue mispricing
"Overdue Phase 3 readouts at Moderna?"pharma_pipeline_catalysts → biotech catalyst calendar
"DoD cybersecurity contracts this week?"usa_award_search → sub-second USAspending mirror
"Median home value and renter share in Lubbock, TX?"housing_market_snapshot + housing_metro_demand
"Unemployment rate last month?"fred_get_series → official FRED data

Cursor picks the right tool via ask_pipeworx — no pack-name memorization required.

How it loads light

The plugin exposes ~26 meta-tools, not all 3,300+ — ask_pipeworx({question}) and friends route at runtime. This matters double in Cursor because Cursor re-sends every tool definition on every turn; a 3,300+-tool dump would dominate the context budget.

Free tier + signup

50 calls/day anonymous, IP-bound. Sign up free in 10s via GitHub for 200/day + a stable account.

Verify after install

Open the MCP settings panel — pipeworx should show a green dot with ~26 tools. Then try in chat:

What was the unemployment rate last month?

What's loaded

  • ask_pipeworx — natural-language router across all 750+ sources.
  • discover_tools — top-20 relevant tools for a task, with full schemas.
  • entity_profile / compare_entities / recent_changes / resolve_entity — fan-out across multiple packs in one call.
  • validate_claim — fact-check claims against SEC XBRL.
  • remember / recall / forget — persistent memory across sessions.
  • list_packs / search_packs / get_pack_tools / get_connection_config / get_platform_status / search_mcp_directory — browse the catalog.

The bundled skill + always-on rule teach Cursor when to reach for each.

Direct pack access

For a specific pack's tools loaded directly (e.g., attom_property_search without going through ask_pipeworx), add a scoped entry to your project's .cursor/mcp.json:

{
  "mcpServers": {
    "pipeworx-attom": {
      "url": "https://gateway.pipeworx.io/attom/mcp"
    }
  }
}

Or a vertical bundle (e.g., ?vertical=housing for the housing-data stack).

Bring your own key

For BYO-tier limits (200/day) or to attach your own per-tool API keys, add an X-API-Key header in mcp.json:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/pipeworx-catalog/mcp",
      "headers": { "X-API-Key": "YOUR_KEY_HERE" }
    }
  }
}

Links

License

MIT


⭐ Star if you'd use this — helps other Cursor users discover it.

Ver o original no GitHub

Bom para

Relacionados